| Abstract Scope |
Multi-laser powder bed fusion (M-LPBF) extends laser powder bed fusion (LPBF) with multiple, independent, simultaneously operating laser sources to overcome the build-rate limitations of single-laser systems, and it is increasingly adopted for industrial-scale metal additive manufacturing (AM). Realizing this productivity gain, however, introduces a new process-planning problem: the slicing strategy used to fill the build area, i.e., the order in which discretized regions of the layer are scanned and the laser assigned to each region, must be chosen from a combinatorially large design space, and a poor choice can produce excessive residual stress buildup. Evaluating a candidate slicing strategy with a high-fidelity, coupled thermo-mechanical simulation is accurate but computationally expensive, often requiring hours to days per case, which makes exhaustive search over the slicing design space impractical. In this work, we present an optimization framework built around a lightweight, reduced-order model (ROM) for predicting the residual stress induced by a candidate slicing strategy, coupled with a genetic algorithm (GA) that optimizes cell scan order and, for multi-laser builds, also laser assignment. The ROM approximates the scan-to-stress relationship through a five-step pipeline: heat memory, shrinkage, settling, and stress mapping, built on an exponentially decaying heat-memory model calibrated against higher-fidelity thermal simulations. We demonstrate the ROM on representative single-and multi-laser examples and apply the coupled ROM-GA framework to a small multi-laser slicing problem on a 3×3 cell grid, reducing the root-mean-square (RMS) von Mises stress objective by 5.92% in approximately 50 minutes of computation. These results indicate that the proposed framework can support fast, first-pass exploration of the slicing design space for multi-laser systems, to be paired with higher-fidelity local refinement and physical validation in future work. |